Within the prior art, echoes within telephone switching systems are well known. Such echoes are normally caused by hybrid unbalanced conditions at a four-wire to two-wire conversion points in a local exchange carrier network or a telephone set or both. In addition, within a telephone set, acoustical feedback can cause echoes as well. There are two basic factors that determine whether echoes are perceived by humans or not. These two factors are highly interrelated. The first factor is the signal level of the echo return signal (also referred to as acoustic to acoustic echo path loss) which is defined as the level of the returned echo signal relative to the transmitted voice signal. The second factor is the time offset of the echo return signal relative to when the voice signal was generated by the talker. FIG. 1 illustrates in graphic form the manner in which loss (manifested in the relative strength of an originating signal and the strength of the returned echo signal) can be utilized to migrate the affects of echo. The lines such as lines 101 and 102 illustrate the echo path delay in milliseconds plotted against the acoustic-to-acoustic echo path loss in dB along the horizontal scale. Plotted on the vertical scale is the rating given by an average group of listeners with the percentage indicating the members of the group who believe that the resulting speech was good or better. The definition and use of the average group is defined in the book entitled Transmission Systems for Communications, Bell Telephone Laboratories, 5th Edition, 1982. Examining line 102, it can be seen that an average group of listeners finds an echo of 5 milliseconds very acceptable if the difference in the echo path loss is in excess of 30 dB. Conversely, if the echo path delay is 5 milliseconds and there is no loss, line 102 shows that only 30 percent of an average group would find this an acceptable telephone conversation. Even for a large echo delay of 1200 milliseconds as illustrated in line 103, if the echo path loss is 60 dB, 90 percent of an average group find that this amount of delay is acceptable. Contrast this against 1.5 milliseconds of echo path delay as illustrated by line 101 with no echo path loss. In this situation, only 70 percent of an average group would find acceptable a delay of 1.5 milliseconds with no echo path loss.
The human perception of echoes verses echo path loss has been well understood within the telephone industry for many years. The designers of prior art telephone switching systems have utilized the manipulation of path loss (referred to as the loss plan technique) to mitigate negative human perception of echoes. The loss plan technique was particularly effective when the national telephone system was controlled by the Bell System. The Bell System was able to implement the loss plan technique effectively. This technique was also aided by the fact that the majority of the prior art telephone switching equipment was circuit-switched equipment or time division multiplex, both of these types of switching systems have low delay times (on the order of a few milliseconds), because of this, the loss plan technique was capable of controlling the perception of echoes.
However, even in prior art switching systems, it has been necessary from time to time to utilize external echo cancellation circuits for severe cases. Indeed, the perceptual effects of echoes due to time offset as well as a high echo return signal are known. When echo returns are high, but delay is low, the perceptual effect is a side tone effect similar to the high side tones experienced in some European countries. On the other hand, the barrel perceptual effect which is encountered when two telephone sets are offhook at the same time occurs from relatively low time offsets in the range of 30–40 msec. When delays in the echo path are long, the perceptual effect is similar to the effect of bouncing ones voice off a mountain.
Echo cancellers (also referred to as echo cancellation circuits) for switching networks are normally finite impulse response digital filters that are implemented using DSP or ASIC circuits. These filters have the advantage that the device resources needed are roughly linearly proportioned to the echo cancellation tail length. An echo cancellation tail length is the time period relative to the reference between the end of the speech burst at the transmitting end and receipt of the end of the echo return at the transmitting end. The cost of an echo canceller is determined to a large extent by the length of the echo cancellation tail for which the echo canceller can compensate. Because the cost of echo cancellers increases as the echo tail length capability increases, it is highly desirable not to utilize echo cancellers that have an echo cancellation tail length greater than what is needed. Another type of echo canceller is an infinite impulse response filter which requires fewer resources than the finite impulse response digital filter but has stability problems.
The prior art telephone switching systems have approached the echo problem in two basic ways. The first is that adopted by the interexchange carriers which is to put an echo canceller on every link going to the local exchange carriers. The second method that has been adopted by most PBX (also referred to as business communications systems or enterprise switching systems) manufacturers has been to add echo cancellers to links to a local exchange carrier only when the need has arisen in the field. The technique utilized by the interexchange carriers is economic for these carriers since their connection to the local exchange carriers is only via high capacity digital trunks. Interexchange carriers deploy echo cancellers at the point of termination between their networks and local exchange carrier networks to avoid having problems with echoes generated in the local exchange networks being perceived by users as an interexchange carrier problem. For a variety of reasons that are described in the following paragraphs, PBX manufacturers are not free because of economic constraints to adopt the method used by the interexchange carrier nor will their prior art technique of adding echo cancellers on a need based scheme work either. A PBX is in many cases placed in the network between a local exchange carrier and an interexchange carrier. A PBX experiences the same echo environment as that seen by an interexchange carrier, and could be indicted by users as causing echo problems which actually occur in local exchange carrier networks. If not dealt with by the PBX, then, these problems are perceived by customers as being problems within the PBX.
PBX and other types of intermediate switch manufacturers face a number of problems with respect to echoes due to the changing environment in which PBXs are being used. The prior art PBX normally connected to telephones that were part of the PBX system (referred to as intercom telephones), local exchange carriers and occasionally to interexchange carriers. However, the prior art PBXs rarely were utilized to communicate a number of calls from a telephone connected to the local exchange carrier to an interexchange carrier. In this case, the PBX resides between the local exchange carrier and the interexchange carrier, and the echo problems of the local exchange carrier are assumed by the customers to be caused by the PBX. Where in reality, the problem is in the local exchange carrier with the delay through the interexchange carrier simply making these echoes perceptually more pronounced. One such situation is where the PBX is used as a call center system and has a number of remote call center agents connected through a local exchange carrier to the PBX. The PBX is receiving “800” type calls from the interexchange carrier and then is re-routing these calls via the local exchange carrier to the remote call center agents. The problem becomes particularly severe where the PBX is interconnected to the local exchange carrier via analog trunks.
Within the prior art it is known to transmit voice calls over a packet switching network such as WAN 213 as illustrated in FIG. 2. The telecommunication terminals utilized to directly connect to a WAN are terminals such as soft phone 214 and 215. A soft phone may be a stand alone piece of equipment that generates and properly addresses packets containing voice information for transmission over WAN 213, or it may be a call application running on a personal computer. WAN 213 can be a combination of various packet switching systems such as ATM, Internet, or Intranet. The problems associated with communicating calls over WAN 213 are the delays that can occur during the transmission. In addition, another problem is that soft phones are implemented by various entities and there are no standards for controlling the acoustic echoes that can occur in a soft phone. Consequently, if a call is established between soft phone 214 and soft phone 215 via WAN 213 and soft phone 215 is not controlling the acoustic echo, the user of soft phone 214 may find the echo objectionable for the reasons that were set out with respect to FIG. 1. Mainly, the echo will be objectionable if the delay through WAN 213 is long, and the echo is of sufficient amplitude. If the call is being switched over the Internet, the delay will be long. Note, if the call is being switched by an ATM network within WAN 213, the delay will not be large. In addition, if soft phone 214 is engaged in a conversation with telephone 224 via WAN 213 and local office 221, this connection may have large echoes occurring because of mismatches in the telephone link between the local office 221 and telephone 224. These large amplitude echoes when combined with a large delay through WAN 213 will make the conversation objectionable to the user of soft phone 214. Similarly, if soft phone 214 is engaged in a conversation with wireless phone 218 via WAN 213 and cellular switching network 216. The resulting echo experienced within cellular switching network 216 in conjunction with its connection to wireless phone 218 may be objectionable to the user of soft phone 214. The prior art technique of introducing loss into the return path of the echo cannot be implemented in WAN 213. Nor, do prior art echo cancellers perform well to stop echoes from being objectionable to the user of soft phone 214. The reason is that prior art echo cancellers require computational resources that are roughly linearly proportioned to the echo cancellation tail length. Because of the large delay that can occur in WAN 213, the echo cancellation tail length becomes extremely long, and a prior art echo canceller is not economically implementable in a soft phone such as soft phone 214. Soft phone 214 tends to be either computationally limited in resources or is a relatively inexpensive device that cannot afford the cost of providing an echo canceller that has a very long tail length.